People count from the crowd using unsupervised learning technique from low resolution surveillance videos
A. Kowcika · 2017
Higher density crowded areas becomes difficult for officials to identify the number of people present within certain area. For the purpose of counting the number of people in an area, image processing concepts are highly supported. Several methods were involved for this counting, but it was capable only to be performed with still images. It is not that much complex to identify the number of people from an image in which it looks visually static. Further numbers of people are counted from the running videos. In this paper we introduce with the identification of people count in a crowd from an ongoing video, in which all the living objects in the video keeps moving. Our proposed work involves with algorithms for background extraction, detection of people, tracking of people and finally counting of the people. We perform background extraction by using ViBE algorithm, then unsupervised learning neural network is executed for detecting people and we use kalman filter for tracking the people. All these advanced algorithms involved here, results with better performances than the state-of-the-art works. Our proposed experimental result obtained by executing our work shows better accuracy results.